FYS 4220 – 2011 / #4
Real Time and Embedded Data Systems and Computing
Message based Process-Process
Synchronization and Communication
Process-to-process messages
• Use of a single construct for both synchronisation and transfer of data
• Three issues:
– the model of message transfer, synchronous vs. asynchronous – the method of process naming
– the message structure
time
Process P1
send message
Process P2
receive message
Synchronous message handling
• No buffering, space for only one message, blocking
• Known as a rendezvous in Ada
– Analogy: a telephone connection, or the HTTP protocol
– A send-receive pair is both data transfer and synchronization
• But how can receiver process know that there is a message for her?
Process P1 Process P2
send
receive
Blocks until receiver is ready to receive the
message ”Hi, there”
Blocks until sender has sent the message
Buffered communication
• Buffer space required
– Buffer write/read operations
– Blocking occurs when buffer is full for write or empty for read
• However, POSIX Asynchronous I/O is non-blocking!
Process P1 Process P2
send message
receive message
blocked
Message buffers
send reply receive reply
blocked
Asynchronous communication
• For example, an application may need to notify another that an event has occurred, but does not need to wait for a response.
• Another example occurs in publish/subscribe
distributed systems, where an application "publishes"
information for any number of clients to read. (A Real-Time example is the CERN ALICE High Level Trigger system comprising ~1000 processors)
• In both these examples it would not make sense for the sender of the information to have to wait if, for example, one of the recipients had crashed.
• However, the communication may include a reply message, in this case the initial sender may want to read the answer when it suits her.
Publisher/subscribers
• Searching the web for this method, one sees that it is mainly described in the context of general,
distributed systems, for instance in servicing customers. http://msdn2.microsoft.com/en-us/library/ms978603.aspx
• However, it is also very interesting for distributed data acquisition / processing configurations
Some problems with asynchronous communication
• Potentially infinite buffers are needed to store unread messages
• The programs get more complex, in particular when signals (to be described in a later lecture) are used to notify the receiver
• It is probably more difficult to prove the correctness of a system
• What to do with messages which seems to be never read?
– In the IEEE 1596 Scalable Coherent Interface one can specify ”time-of-death” for a message
Process naming
• Two distinct sub-issues
– direction versus indirection – symmetry
• With direct naming, the sender explicitly names the receiver:
send <message> to <process-name>
• With indirect naming, the sender names an
intermediate entity (e.g. a channel, mailbox, link or pipe):
send <message> to <message queue>/channel/mailbox
(ref. B&W)
Process naming
(contd)• A naming scheme is symmetric if both sender and receiver name each other (directly or indirectly)
send <message> to <process-name>
wait <message> from <process-name>
send <message> to <mailbox>
wait <message> from <mailbox>
• It is asymmetric if the receiver names no specific source but accepts messages from any process
wait <message>
• Asymmetric naming fits the client-server paradigm
• With indirect the intermediary could have:
– a many-to-one structure – a many-to-many structure – a one-to-one structure
– a one-to-many (as in publisher/subscribers)
(ref B&W)
”Message queues”
• The generic name Message Queues is a software- engineering component used for interprocess
communication or inter-thread communication within the same process. It uses a queue for messaging – the passing of control or of content
• Message queues provide an asynchronous
communications protocol, meaning that the sender and receiver of the message do not need to interact with the message queue at the same time.
Messages placed onto the queue are stored until the receiver retrieves them.
VxWorks message queues
• Modern real-time applications are constructed as a set of
independent but cooperating tasks. While semaphores provide a high-speed mechanism for the synchronization and
interlocking of tasks, often a higher-level mechanism is
necessary to allow cooperating tasks to communicate with each other. In VxWorks, the primary intertask communication
mechanism within a single CPU is message queues. This is also seen in WindView traces.
• Message queues allow a variable number of messages, each of variable length, to be queued. Any task or Interrupt Service
Routine (ISR – more about that in a later lecture) can send messages to a message queue. Any task can receive
messages from a message queue. Multiple tasks can send to and receive from the same message queue. Full-duplex
communication between two tasks generally requires two message queues, one for each direction
”Hello World” – intertask communication
#include <iostream.h>
using namespace std;
void main() {
cout << "Hello World!" << endl;
cout << "Welcome to C++ Programming" << endl;
}
WindView trace of ”Hello World”, see next page, note the message queue operations
WindView trace of Hello World
VxWorks Wind and POSIX message queues
• There are two message-queue subroutine libraries in VxWorks
• The first of these, msgQLib, provides Wind
message queues, designed expressly for VxWorks;
• the second, mqPxLib, is compatible with the POSIX standard (1003.1b) for real-time extensions
VxWorks Wind message queues
• Wind message queues are created and deleted with the routines shown below. This library provides
messages that are queued in FIFO order, with a single exception: there are two priority levels, and messages marked as high priority are attached to the head of the queue.
– msgQCreate( ) Allocate and initialize a message queue – msgQDelete( ) Terminate and free a message queue.
– msgQSend( ) Send a message to a message queue
– msgQReceive( ) Receive a message from a message queue
VxWorks Wind create message queue
• A message queue is created with msgQCreate( ). Its parameters specify the maximum number of
messages that can be queued in the message queue and the maximum length in bytes of each message. Enough buffer space is preallocated for the specified number and length of messages
MSG_Q_ID msgQCreate (
int maxMsgs, /* max messages that can be queued */
int maxMsgLength, /* max bytes in a message */
int options /* message queue options */
)
options:
MSG_Q_FIFO (0x00) queue pended tasks in FIFO order MSG_Q_PRIORITY (0x01) queue pended tasks in priority order
VxWorks Wind msg send message
• A task or ISR sends a message to a message queue with msgQSend( ). If no tasks are waiting for
messages on that queue, the message is added to the queue's buffer of messages. If any tasks are already waiting for a message from that message queue, the message is immediately delivered to the first waiting task
STATUS msgQSend (
MSG_Q_ID msgQId, /* message queue on which to send */
char * buffer, /* message to send */
UINT nBytes, /* length of message */
int timeout, /* ticks to wait */
int priority /* MSG_PRI_NORMAL or MSG_PRI_URGENT */
)
VxWorks Wind msg send priorities
• The msgQSend( ) function allows specification of the priority of the message as either
normal (MSG_PRI_NORMAL) or urgent (MSG_PRI_URGENT).
Normal priority messages are added to the tail of the list of queued messages, while urgent priority
messages are added to the head of the list
VxWorks Wind receive message
• A task receives a message from a message queue with msgQReceive( ). If messages are already available in the message queue's buffer, the first message is immediately dequeued and returned to the caller. If no messages are
available, then the calling task blocks and is added to a queue of tasks waiting for messages. This queue of waiting tasks can be ordered either by task priority or FIFO, as specified in an option parameter when the queue is created
int msgQReceive (
MSG_Q_ID msgQId, /* message queue from which to receive */
char * buffer, /* buffer to receive message */
UINT maxNBytes, /* length of buffer */
int timeout /* ticks to wait,
or NO_WAIT (0) or WAIT_FOREVER (-1) */
)
VxWorks Wind msg timeouts
• Both msgQSend( ) and msgQReceive( ) take
timeout parameters. When sending a message, the timeout specifies how many ticks to wait for buffer space to become available, if no space is available to queue the message. When receiving a message, the timeout specifies how many ticks to wait for a message to become available, if no message is immediately available. As with semaphores, the
value of the timeout parameter can have the special values of NO_WAIT (0), meaning always return
immediately, or WAIT_FOREVER (-1), meaning never time out the routine
Example 2.8 VxWorks Programmer’s Guide
Wind message queue benchmark
• Below is the result of a simple message queue benchmark in
kB/sec for VxSim on a DELL D420 laptop with 1.20 GHz Intel CPU
POSIX message queues
(text: ref B&W)• POSIX supports asynchronous, indirect message passing through the notion of message queues
• A message queue can have many readers and many writers
• Priority may be associated with the queue
• Intended for communication between processes (not threads)
• Message queues have attributes which indicate their maximum size, the size of each message, the
number of messages currently queued etc.
• An attribute object is used to set the queue attributes when the queue is created
POSIX message queues
• Message queues are given a name when they are created
• To gain access to the queue, requires an mq_open name
• mq_open is used to both create and open an already existing queue (also mq_close and mq_unlink)
• Sending and receiving messages is done via mq_send and mq_receive
• Data is read/written from/to a character buffer.
• If the buffer is full or empty, the sending/receiving process is blocked unless the attribute O_NONBLOCK has been set for the queue (in which case an error return is given)
• If senders and receivers are waiting when a message queue becomes unblocked, it is not specified which one is woken up unless the priority scheduling option is specified
POSIX message queues
• A process can also indicate that a signal should be sent to it when an empty queue receives a message and there are no waiting receivers
• In this way, a process can continue executing whilst waiting for messages to arrive or one or more message queues. This concept, named Asynchronous I/O or non- blocking I/O, will be discussed in a later lecture.
• It is also possible for a process to wait for a signal to arrive; this allows the equivalent of selective waiting to be implemented
• If the process is multi-threaded, each thread is considered to be a potential sender/receiver in its own right
VxWorks POSIX message queues
• These routines are similar to Wind message queues, except that POSIX message queues provide named queues and messages with a range of priorities. A process can also indicate that a signal should be sent to it when an empty queue receives a message and there are no waiting receivers. Signals will be covered in a later lecture.
• Before a set of tasks can communicate through a POSIX message queue, one of the tasks must create the message queue by calling mq_open( ) with the O_CREAT flag set. Once a message queue is created, other tasks can open that queue by name to send and receive messages on it. Only the first task opens the queue with the O_CREAT flag; subsequent tasks can open the queue for receiving only (O_RDONLY), sending only (O_WRONLY), or both sending and receiving (O_RDWR)
VxWorks POSIX mq routines
• mqPxLibInit( ) Initialize the POSIX message queue library (non-POSIX)
mq_open( ) Open a message queue.
mq_close( ) Close a message queue.
mq_unlink( ) Remove a message queue.
mq_send( ) Send a message to a queue.
mq_receive( ) Get a message from a queue.
mq_notify( ) Signal a task that a message is waiting on a queue.
mq_setattr( ) Set a queue attribute.
mq_getattr( ) Get a queue attribute.
VxWorks POSIX mq_send( )
• To put messages on a queue, use mq_send( ). If a task attempts to put a message on the queue when the queue is full, the task blocks until some other task reads a message from the queue, making space available. To avoid blocking on mq_send( ), set O_NONBLOCK when you open the message queue. In that case, when the queue is full, mq_send( ) returns -1 and sets errno to EAGAIN instead of pending, allowing you to try again or take other action as appropriate
• One of the arguments to mq_send( ) specifies a message priority. Priorities range from 0 (lowest priority) to 31 (highest priority).
VxWorks POSIX mq_receive( )
• When a task receives a message using mq_receive( ), the task receives the highest-priority message currently on the queue. Among multiple messages with the same priority, the first message placed on the queue is the first received (FIFO order). If the queue is empty, the task blocks until a message is placed on the queue. To avoid pending on mq_receive( ), open the message queue with O_NONBLOCK; in that case, when a task attempts to read from an empty queue, mq_receive( ) returns -1 and sets errno to EAGAIN
Example 2.9 VxWorks Programmer’s Guide –p1
Example 2.9 VxWorks Programmer’s Guide –p2
Example 2.9 VxWorks Programmer’s Guide –p3
VxWorks demo p 1 of 4
/* VxWorks Wind and POSIX message queues */
#include "vxWorks.h"
#include "msgQLib.h"
#include "taskLib.h"
#include "time.h"
#include "stdio.h"
#include "POSIXerrno.c“ /* Skaali routine */
/* Number of messages in queue, length of message */
#define MSG_MAX 4
#define MSG_LEN 16
MSG_Q_ID myMsgQId;
STATUS show_OBJ_error (int errn) {
if (errn == S_objLib_OBJ_ID_ERROR) printf("OBJ_ID_ERROR\n");
if (errn == S_objLib_OBJ_UNAVAILABLE) printf("OBJ_UNAVAILABLE\n");
if (errn == S_objLib_OBJ_DELETED) printf("OBJ_DELETED\n");
if (errn == S_objLib_OBJ_TIMEOUT) printf("OBJ_TIMEOUT\n");
if (errn == S_objLib_OBJ_NO_METHOD) printf("OBJ_NO_METHOD\n");
return (OK);
}
VxWorks demo p 2 of 4
/* --- WIND message queue example --- */
STATUS msqWIND (void) {
int msgl, loop;
char msgBuf[MSG_LEN];
msgl = MSG_LEN; /* max bytes in a message */
/* create message queue */
if ((myMsgQId = msgQCreate (MSG_MAX, MSG_LEN, MSG_Q_PRIORITY)) == NULL) return (ERROR);
/* fill message queue */
for (loop = 0; loop<MSG_MAX; loop++) { if (msgQSend (myMsgQId, msgBuf, msgl,
WAIT_FOREVER, MSG_PRI_NORMAL) == ERROR) return (errno);
}
printf("fill message queue OK, try one more message in NO_WAIT mode\n");
if (msgQSend (myMsgQId, msgBuf, msgl,
NO_WAIT, MSG_PRI_NORMAL) == ERROR) { show_OBJ_error (errno);
return (errno);
}
msgQDelete (myMsgQId);
return (OK);
}
VxWorks demo p 3 of 4
/* --- POSIX message queue example --- */
#include "mqueue.h"
#include "fcntl.h"
#define MSQ_NAME "POSIXmsq"
#define MODE 0
#define MSGPRI 31
STATUS msqPOSIX() {
mqd_t mq_descr; /* msg queue descriptor */
mode_t mode=(mode_t)MODE; /* mode, dummy */
struct mq_attr ma; /* queue attributes */
int loop;
char msgBuf[MSG_LEN];
/* set the required message queue attributes */
ma.mq_flags = 0;
ma.mq_flags = O_NONBLOCK; /* Immediate return */
ma.mq_maxmsg = MSG_MAX;
ma.mq_msgsize = MSG_LEN; /* max bytes in a message */
/* Create and open message queues */
if ((mq_descr = mq_open (MSQ_NAME, O_RDWR|O_CREAT, mode, &ma))
== (mqd_t)-1) { return(errno);
}
/* fill message queue */
for (loop = 0; loop<MSG_MAX; loop++)
if (mq_send (mq_descr, msgBuf, ma.mq_msgsize, MSGPRI) != OK) { printf("\nmq_send error %d", errno);
mq_unlink (MSQ_NAME);
}
printf("fill message queue OK, try one more message in NONBLOCK mode\n");
if (mq_send (mq_descr, msgBuf, ma.mq_msgsize, MSGPRI) != OK) { printPOSIXerrorcode (errno);
mq_unlink (MSQ_NAME);
return (errno);
}
return (OK);
VxWorks demo p 4 of 4
Run the message queue demo tasks:
-> msqPOSIX
fill message queue OK, try one more message in NONBLOCK mode POSIX error EAGAIN - No more processes
value = 2 = 0x2 -> msqWIND
fill message queue OK, try one more message in NO_WAIT mode OBJ_UNAVAILABLE
value = 3997698 = 0x3d0002 ->
Note different error return codes from Wind and POSIX send-message when message buffer is full!
Summary of VxWorks message queue features
Message Queue Feature Comparison
Feature Wind Message Queues POSIX Message Queues
Message Priority Levels 1 32
Blocked Task Queues FIFO or priority-based Priority-based Receive with Timeout Optional Not available Task Notification Not available Optional (one task)
Close/Unlink Semantics No Yes
Another important feature of POSIX message queues is, of course, portability: if you are migrating to VxWorks from another 1003.1b-compliant system, using POSIX message queues enables you to leave that part of the code unchanged, reducing the porting effort.
Servers and Clients with Message Queues
Real-time systems are often structured using a client-server model of tasks. In this model, server tasks accept requests from client tasks to perform some service, and usually return a reply. The requests and replies are usually made in the
form of intertask –(process) messages. In VxWorks, message queues or pipes are a natural way to implement this.
For example, client-server communications might be
implemented as shown in figure. Each server task creates a message queue to receive request messages from clients.
Each client task creates a message queue to receive reply messages from servers. Each request message includes a field containing the msgQId of the client's reply message queue. A server task's "main loop" consists of reading request messages from its request message queue,
performing the request, and sending a reply to the client's reply message queue.
Other I
nterP
rocessC
ommunicationfacilities
• Pipes- VxWorks supports pipe devices
• status = pipeDevCreate ("/pipe/name", max_msgs, max_length);
• Tasks can use standard I/O routines to open, read and write pipes, and invoke ioctl routines
• Sockets
• A socket is an endpoint for communications between tasks; data is sent from one socket to another.
• VxWorks supports the Internet protocols TCP and UDP
• Remote Procedure calls
• Remote Procedure Calls (RPC) is a facility that allows a process on one machine to call a procedure that is executed by another process on either the same machine or a remote machine. Internally, RPC uses sockets as the underlying communication mechanism. Thus with RPC, VxWorks tasks and host system processes can invoke routines that execute on other VxWorks or host machines, in any combination.
• Some IPC methods will be discussed further in the lecture on Real-Time facilities and I/O. IPC is a key issue in Computer science, and for those who are interested, a few zillion papers are available on this subject